Was running Unraid fine with only 8GB of RAM, using only Plex docker. Wife complained that she couldn't watch a DVR'd show through plex. Turns out she was trying to watch it while it was still recording, i.e. the recording started at 3PM but she started watching from the beginning at ~3:30PM and when trying to advance through commercials it would choke. So I figured I'd order some RAM and throw it in there. Added two 8GB modules to now total 24GB of RAM. Booted up, ran memtest for 5+ hours with no errors, then rebooted. 

Everything appears to work fine but my CPU cores keep getting maxed out, one at a time - the bar graph on the Dashboard is constantly fluxuating between cores. It's not the plex docker - I disabled Docker and rebooted and problem persists. I'm running an Nvidia card and using the Nvidia driver plugin. It is set to the latest driver version. When I open terminal and view top, it shows nv_open_q varying between 50 and 100% CPU usage. 

 

Diagnostics attached. Any help is greatly appreciated. I'm not as tech savvy as most so go easy on me : )

unraid-diagnostics-20241113-1023.zip

the watch while recording is a known issue and is a g-card transcode thing, not ram.

Running Plex and how it decided to encode to watch is what would be causing random cpu hangup as a resource load.

 

in docker, pin it to xyz cores to force those in use for its loaded resources. Confirm plex setting is off auto and set to the g-card
 

 

Edited November 13, 20241 yr by bmartino1
typo - data

Understood about the plex dvr issue - thanks. But problem remains regarding the high cpu usage. 

 

I never had this cpu spiking when I was running just 8GB (2x4GB) of RAM. Only thing I changed was I added 16GB (2x8) to now total 24GB of RAM. Upon reboot, CPU is constantly spiking. Two processes were spiking it. One, shfs, is now resolved (see below). nv_open_q is still causing the issue.

shfs: this process was utilizing quite a bit of CPU as well. I found a mention on reddit to have plex mount /mnt/cache/appdata instead of /mnt/user/appdata and that solved that one.

nv_open_q: this one was still utilizing a lot of CPU and spiking it. Uninstalled nvidia driver completely, rebooted, checked the stats to make sure no cpu usage spikes (confirmed); reinstalled nvidia driver - same issue with nv_open_q spiking CPU intermittently.

 

Changed from Production branch to Latest branch. Same issue. Changed from Latest branch to New Feature branch. Rebooted after each of course. Any ideas?

in terms of the gcard NV)open_q. no you would need to ask ICH777on the nvdia support forum.

Outside of docker pinning to tell the docker it gets to use xyz many cores, i don't have any other solution.